Background:Deep learning models showed great success and potential when applied to many biomedical problems. However, the accuracy of deep learning models for many disease prediction problems is affected by time-varying covariates, rare incidence, and covariate imbalance when using structured electronic health records data. The situation is further exasperated when predicting the risk of one disease on condition of another disease, such as the hepatocellular carcinoma risk among patients with nonalcoholic fatty liver disease due to slow, chronic progression, the scarce of data with both disease conditions and the sex bias of the diseases. Objective:The goal of this study is to investigate the extent to which time-varying covariates, rare incidence, and covariate imbalance influence deep learning performance, and then devised strategies to tackle these challenges. These strategies were applied to improve hepatocellular carcinoma risk prediction among patients with nonalcoholic fatty liver disease. Methods:We evaluated two representative deep learning models in the task of predicting the occurrence of hepatocellular carcinoma in a cohort of patients with nonalcoholic fatty liver disease (n = 220,838) from a national EHR database. The disease prediction task was carefully formulated as a classification problem while taking censorship and the length of follow-up into consideration. Results:We developed a novel backward masking scheme to evaluate how the length of longitudinal information after the index date affects disease prediction. We observed that modeling time-varying covariates improved the performance of the algorithms and transfer learning mitigated reduced performance caused by the lack of data. In addition, covariate imbalance, such as sex bias in data impaired performance. Deep learning models trained on one sex and evaluated in the other sex showed reduced performance, indicating the importance of assessing covariate imbalance while preparing data for model training. Conclusions:Devising proper strategies to address challenges from time-varying covariates, lack of data, and covariate imbalance can be key to counteracting data bias and accurately predicting disease occurrence using deep learning models. The novel strategies developed in this work can significantly improve the performance of hepatocellular carcinoma risk prediction among patients with nonalcoholic fatty liver disease. Furthermore, our novel strategies can be generalized to apply to other disease risk predictions using structured electronic health records, especially for disease risks on condition of another disease.

Abstract Introduction Low testosterone has been suggested as a contributor to mortality and rejection after solid organ transplant. Whether testosterone therapy specifically improves these outcomes is unknown. Objective To determine whether treatment of low testosterone with testosterone therapy is associated with improved rejection or mortality after solid organ transplant using TriNetX, a large deidentified, multi-institutional database. Methods We compiled all clinical data from the TriNetX research network, a globally federated health research network (with waiver from Western IRB) that provides de-identified clinical information from 58 heath care organizations, and over 80 million patients located within the United States. We selected male patients (aged ≥ 18) with a history of lung, kidney, liver, heart, or pancreas transplant. We categorized patients into two groups, based on prescription for testosterone therapy. For all patients, we collected clinical information including demographics, comorbidities, laboratory findings, and medication use. Our outcomes of interest were 1-year, 3-year, and all-time mortality, as well as allograft rejection. We evaluated differences in baseline characteristics between testosterone level (Normal T, or Low T), using chi-square or Fisher’s exact test for categorical data (presented as frequencies and percentages), and independent t-test for continuous data. To address potential confounders that could bias our results, we used a propensity score weighted regression (inverse probability of treatment weighting), matched on age, race/ethnicity, comorbid conditions, multiple transplant status, and immunosuppressant use. Multivariable logistic regression was used to model outcome of transplant rejection, on the predictors of testosterone level, testosterone therapy, and their interaction, while a cox regression model was used to determine probability of mortality. Results A total of 1,762 men were identified with solid organ transplants. 262 (14.9 %) met criteria for low testosterone and had a history a prescription of testosterone therapy. Baseline characteristics were similar between the two groups (Table 1). After matching for existing patient characteristics, outcomes showed that a prescription for testosterone therapy was associated with improved mortality after organ transplant. The risk of mortality was reduced at 1 year (HR 0.33), 3 years (HR0.57) and for all time mortality (HR 0.61) among patients treated with testosterone therapy. No effect of testosterone treatment was seen with regards to allograft rejection (data not shown). Conclusions Testosterone therapy may improve mortality in men with low serum testosterone following solid organ transplant. Further studies are required to determine when and what patients derive most benefit from testosterone therapies, particularly as the result does not appear to be driven by reduced allograft rejection. Disclosure No

Abstract Introduction Low serum testosterone has previously been suggested as a predisposing factor contributing to mortality after solid organ transplant. Previous studies on this topic stem from single-center retrospective series. Objective To determine the impact of low serum testosterone as a risk factor for mortality after solid organ transplantation using TriNetX, a large de-identified, multi-institutional database. Methods Data source, Patient Selection, and Outcomes We compiled all clinical data from the TriNetX research network, a health research network (with waiver from Western IRB) that provides de-identified clinical information from 58 health care organizations, and over 80 million patients located within the United States. We selected male patients (age ≥ 18) with a history of lung, kidney, liver, heart, or pancreas transplant. Cases were included if serum testosterone was drawn within one year before transplant. We excluded all patients previously on testosterone therapy. We categorized patients into two groups, based on testosterone levels. We defined patients as having low testosterone (Low T) by the following criteria: total testosterone < 300 (ng/Dl), or a diagnosis of hypogonadism (ICD-10: E29.1). For all patients, clinical characteristics/demographics including comorbidities, laboratory findings, and medication use were compared using descriptive statistics. Our outcomes of interest were mortality at 1-year, 3-year, or at any time. Statistical analyses We evaluated differences in baseline characteristics between testosterone levels (Normal T, or Low T), using chi-square or Fisher’s exact test for categorical data and an independent t-test for continuous data. To address potential confounders that could bias our results, we used a propensity score weighted regression matched on age, race/ethnicity, comorbid conditions, multiple transplant status, and immunosuppressant use. A Cox regression model was used to determine the probability of mortality. All analyses were performed using R version 4.0.4. Results A total of 1,762 men were identified with solid organ transplants. Of these, 792 men met criteria for low testosterone (Table 1). Low testosterone correlated with higher rates of mortality at 1 year, 3 years, and all-time mortality (Figure 1). Conclusions Low testosterone at the time of transplant is correlated with mortality up to 3 years after organ transplantation. Future work will focus on individual organ transplants and whether modifying this risk factor with testosterone therapy changes outcomes. Disclosure No

Abstract Introduction Men with end stage organ dysfunction or failure are more likely to have low testosterone. In the case of men with heart and kidney failure, improvement in testosterone levels are seen after successful transplantation. Limited data exists regarding the natural course of serum testosterone after lung transplant. Objective To report incidence of low testosterone in men prior to lung transplant and to determine changes in testosterone levels after transplantation. Methods We compiled all clinical data from the TriNetX research network, a globally federated health research network (with waiver from Western IRB) that provides de-identified clinical information from 58 heath care organizations, and over 80 million patients located within the United States. We selected male patients (aged ≥ 18) with a history of lung transplant and testosterone levels recorded before and after transplantation. All men who had a recorded prescription for testosterone therapy were excluded. For all patients, we collected clinical information including demographics, comorbidities, laboratory findings, and medication use. Patients were stratified by testosterone levels and grouped into those who had low serum testosterone versus those with normal levels, with a cut-off point of 300 ng/dL. Patient characteristics were compared using descriptive statistics. Propensity score matching was performed between the two groups to normalize differences in pre-operative characteristics. Mean serum testosterone levels before and after transplantation was compared using a paired T test to determine change in serum testosterone levels after successful transplantation. Results A total of 37 men were identified with history of lung transplant, and testosterone levels recorded before and after transplant. Average serum testosterone level in the year prior to lung transplantation was 301.57 +/− 163.25. After transplantation, average serum testosterone was 332.37 +/− 200.92, for a mean difference of 30.80 (p = 0.4815). Prior to transplantation, 24 (65%) men met criteria for low serum testosterone. After transplantation 17 (46%) men demonstrated serum testosterone less than 300 ng/dl. Among men with low serum testosterone, average testosterone levels in the year prior to lung transplantation was 203.29 +/− 66.57, and was 332.37 +/− 200.92 after transplantation (p = 0.0057). Conclusions Among men with low serum testosterone receiving lung transplant, transplantation was associated with rescue of low testosterone levels. Further research incorporating a larger cohort is required to determine which patients recover endocrine function after lung transplant and whether etiology of chronic lung disease results in different outcomes after transplantation with regards to serum androgen level and function. Disclosure No

Purpose Diaphragmatic dysfunction is documented after lung transplantation and can affect up to 62% patients. Diaphragm pacing (DP) prevents ventilator induced diaphragm dysfunction (VIDD) while on mechanical ventilation (MV) and has been shown via functional electrical stimulation to improve phrenic nerve recovery. We report the largest experience of DP in lung transplantation recipients. Methods This is a retrospective analysis of an IRB approved prospective, non-randomized interventional experience at a single institution with two DP systems. A chronic DP system [NeuRx, Synapse Biomedical] was implanted laparoscopically in those with phrenic nerve injury or difficulty with weaning from MV remotely after their transplant. A temporary DP system [TransAeris, Synapse Biomedical] was implanted at the time of transplantation or laparoscopically. In both types of implantation, diaphragm stimulation ensued if needed to wean from MV or for nerve recovery. Results evaluated radiographically and with diaphragm electromyography (dEMG). Results DP was utilized in16 patients with no device adverse events. Of those, 5 patients had chronic DP system: a) 1 patient one year post transplant shows no recovery of phrenic nerve injury, still pacing; b) 3 patients showed recovery of phrenic nerve/diaphragm function through pacing; c) 1 patient 2 years post-transplant sustained hip fracture requiring surgery, developed pneumonia and became tracheostomy MV dependent; DP allowed complete diaphragm recovery, decanullation of tracheostomy and removal of DP wires. 11 patients had temporary DP electrodes placed: a) 1 recipient(two months post-transplant) had DP use during ECMO for COVID- 19 sepsis and respiratory failure and subsequently expired when family withdrew therapy; b) 10 implanted at time of lung transplant. Of those 10, 3 patients had bilateral dEMG identified post-operatively with uneventful recovery and removal of electrodes; 7 patients had diaphragm abnormalities identified post-operatively and underwent DP. Of those 7, 5 showed recovery and DP electrodes removed and 2 are still pacing 1 and 9 months post-transplant. Conclusion DP was safely used in lung transplantation to identify and improve recovery of phrenic nerve injuries, wean from MV and prevent VIDD. DP shows promise in addressing diaphragm dysfunction after lung transplantation and improving outcomes. Diaphragmatic dysfunction is documented after lung transplantation and can affect up to 62% patients. Diaphragm pacing (DP) prevents ventilator induced diaphragm dysfunction (VIDD) while on mechanical ventilation (MV) and has been shown via functional electrical stimulation to improve phrenic nerve recovery. We report the largest experience of DP in lung transplantation recipients. This is a retrospective analysis of an IRB approved prospective, non-randomized interventional experience at a single institution with two DP systems. A chronic DP system [NeuRx, Synapse Biomedical] was implanted laparoscopically in those with phrenic nerve injury or difficulty with weaning from MV remotely after their transplant. A temporary DP system [TransAeris, Synapse Biomedical] was implanted at the time of transplantation or laparoscopically. In both types of implantation, diaphragm stimulation ensued if needed to wean from MV or for nerve recovery. Results evaluated radiographically and with diaphragm electromyography (dEMG). DP was utilized in16 patients with no device adverse events. Of those, 5 patients had chronic DP system: a) 1 patient one year post transplant shows no recovery of phrenic nerve injury, still pacing; b) 3 patients showed recovery of phrenic nerve/diaphragm function through pacing; c) 1 patient 2 years post-transplant sustained hip fracture requiring surgery, developed pneumonia and became tracheostomy MV dependent; DP allowed complete diaphragm recovery, decanullation of tracheostomy and removal of DP wires. 11 patients had temporary DP electrodes placed: a) 1 recipient(two months post-transplant) had DP use during ECMO for COVID- 19 sepsis and respiratory failure and subsequently expired when family withdrew therapy; b) 10 implanted at time of lung transplant. Of those 10, 3 patients had bilateral dEMG identified post-operatively with uneventful recovery and removal of electrodes; 7 patients had diaphragm abnormalities identified post-operatively and underwent DP. Of those 7, 5 showed recovery and DP electrodes removed and 2 are still pacing 1 and 9 months post-transplant. DP was safely used in lung transplantation to identify and improve recovery of phrenic nerve injuries, wean from MV and prevent VIDD. DP shows promise in addressing diaphragm dysfunction after lung transplantation and improving outcomes.

Background and aims: Hepatocellular carcinoma (HCC) is a leading cause of cancer mortality. Operative management of early disease includes ablation, resection, and transplantation. We compared the operative management of early-stage HCC in patients by race. Methods: We utilized data from the National Cancer Database (2004-2016) among patients with cT1 HCC and Charlson-Deyo score 0-1 (n=25,029). We used multivariable logistic regression models to compare operative management (none, ablation, resection, and transplantation) by race, adjusted for demographic and clinical factors. We further utilized marginal standardization to estimate adjusted proportions of resection or transplantation, separately, for age and insurance status by race. Results: A total of 25,029 patients were included (White=20,410; Black=4,619). After adjusting for clinicodemographic variables, Black race was associated with a lower likelihood of undergoing operative intervention (OR 0.89, p=0.009). When stratified by type of operative intervention, Black patients were more likely to undergo resection (OR 1.23, p<0.001) and less likely to undergo transplantation (OR 0.60, p<0.001). Marginal standardization models demonstrated Black race was associated with increased probability of resection in patients >50yrs, with private insurance/Medicare (Figure A and B), and lower probability of transplantation regardless of age or insurance payor (Figure C and D). Conclusions: Black race is associated with lower rates of hepatic transplantation and higher rates of hepatic resection for early HCC regardless of age or insurance payor. The etiology of these disparities is multifactorial and correcting the root causes represents a critical area for improvement.